Object Agreement in Hungarian Elizabeth Coppock LFG04, Christchurch 1 As any Hungarian grammar will tell you, indefinite paradigm (present tense) singular plural 1 -Vk -unk/-ünk 2 -sz/-vl -tok/-tek/-tök 3 - -nak/nek definite paradigm (present tense) singular plural 1 -Vm -juk/-jük 2 -Vd -játok/-itek 3 -ja/i -ják/-ik Indefinite object indefinite V: (1) Lát-ok egy madar-at. see-1.sg.indef an bird-acc I see an bird V intransitive V indefinite: (2) Vár-ok. wait-1sg.indef I m waiting. Local object indefinite V: (3) Lát-ott eng-em/ tég-ed/ mi-nk-et/ ti-tek-et see-3sg.indef me-poss:1sg you-poss:2sg 1pl-poss:1pl-acc 2pl-poss:2pl-acc He saw me/you/us/y all. Special conjugation for 2nd person object, 1st singular subject: (4) Lát-l-ak téged. see-2-1sg you.sg.acc I see you (5) Object Agreement Rule (traditional) If the object is second person and the subject is first person singular, then use the -lak/-lek form. If the verb has a definite third person object, then conjugate the verb using the definite paradigm. Otherwise use the indefinite paradigm. 1
Formalization of traditional characterization (6) -Vm V infl ( obj def= c +) (7) -Vk V infl ( obj def) (8) János D ( def)= + (9) az D ( def)= + 2 Accusative Focus-raised Subjects Problem with (5): Accusative focus-raised subjects agree with matrix verb, yet are not objects of that verb. Focus raising construction: (10) Holnap i mond-ott, [hogy jön t i ]. tomorrow say-past.3sg that come.3sg He said that he would come tomorrow Case switch (nom acc) in subject focus-raising: (11) Péter-t i mond-t-a, [hogy t i jön] Péter-acc say-past-3sg.def that come.3sg It is Peter who he/she said is coming. Péter would be nominative downstairs: (12) Evá-nak mond-t-a, [hogy Péter jön] Eve-dat say-past-3sg.def that Peter.nom come.3sg He/she said to Eva that Peter is coming Accusative focus-raised subjects agree in definiteness with matrix V: (13) Két fiu-t i mond-ott [hogy t i jön] two boy-acc say-3sg.indef.past that come.3sg It was two boys that he/she said were coming. Show person agreement too (den Dikken, 1999): (14) Téged i akar-l-ak, [hogy t i jössz] you.acc want-2-1sg comp come.2sg.indef I want you to come. Accusative focus-raised subjects therefore show all coding properties of objects. But do they show behavioral properties of objects, too? 2
2.1 They don t behave as thematic objects Why accusative focus-raised subjects are not arguments of matrix V: 1. Focus-raising doesn t change truth conditions, except for exhaustivity. 2. All bridge verbs would then have two argument structures, one bivalent for the usual case, and one trivalent for focus-raising. The extra argument would only ever surface in the focus position, and it would always long-distance control a lower argument. 3. They don t incorporate. Hungarian noun incorporation: incorporation as juxtaposition (Mithun, 1984) (15) Az orvos beteg-et viszgál-t. the doctor patient-acc examine-past.3sg.indef The doctor patient-examined. Incorporation possible with verbs that take a CP and a genuine direct object: (16) Nem szabad tolmács-ot kér-ni, [hogy leforditsa]. not allowed translator-acc request-inf that translate.sbj It is forbidden to ask a translator to translate. cf. (17) A tolmács-ot kér-t-em, [hogy lefordit-s-a a cikk-et] the translator-acc ask-past-1sg.def comp translate-sbj-3sg.def the article I asked the translator to translate the article Not possible with focus-raising: (18) * Nem szabad nő-t i mond-ani, [hogy t i csunya]. not allowed woman-acc say-inf that ugly It is forbidden to say that women are ugly. Against alternative explanations for the ungrammaticality of (18): Incorporation is possible with any verb: in Hungarian the incoporation construction is freely available, and is not restricted to a set of lexical verbs ( incorporating verbs, like in West Greenlandic for instance). (Farkas and de Swart, 2003, p. 96). There s no problem focussing incorporated nominals: (19) Mari nem ujság-ot olvas, hanem sztori-t. Mary nem newspaper-acc read.3sg.indef but story-acc Mary s not reading a newspaper, but a story. Focus-raised subjects are not arguments of the matrix V. 3
2.2 They don t behave like objects at all Reason for doubt If accusative focus-raised subjects are matrix objects, then this subtype of long-distance focus-movement is control. But control is finite clause-bound. Problem from LFG standpoint: If the construction were control, then the matrix verb would need a constraining equation linking matrix accusative to embedded subject. This puts the burden of constraining the extraction path on the verb, rather than the aspect of grammar that regulates long-distance dependencies. (Coppock, 2003) Argument 1: Binding Normal objects can t be non-reflexively bound by local subject ((21)): (20) Péter i önmagá-t i utálja Peter.nom himself-acc hate-3sgdef Peter hates himself. (21) * Péter i [ö magá-t i ] utálja Peter.nom him himself-acc hate-3sgdef Peter hates himself. Non-reflexive binding of accusative focus-raised subjects by matrix subject is possible, though: (22) Önmagá-t i mondta Péter i hogy szeret-i Mari-t himself-acc say-past-3sgdef Peter.nom that love-3sg.def Mary-acc It is himself i that Peter i said loves Mary. (23) Ő i magá-t mondta Péter i hogy szeret-i Mari-t He himself-acc say-past-3sgdef Peter.nom that love-3sg.def Mary-acc It is him i, himself, that Peter i said loves Mary. Argument 2: Depictives Pre-clausal depictive can modify an accusative object: (24) János-t j kér-t-em i meg részeg-en i/j, [hogy jöjjön] John-acc ask-past-1sg.def perf drunk-ly comp come.3sg.sbj I asked John drunk to come... (és az-ért nem emlékez-ett). and that-because not remember-3sg.indef.past... (and so he didn t remember). But not an accusative, focus-raised subject: (25) János-t j mond-t-am i részeg-en i/ j hogy beszél-get-ett Mari-val. John-acc say-past-1sg drunk-ly that speak-freq-past Mary-with John, I said drunk that he was chatting with Mary. 4
Argument 3: there s already an obj there Berman (2003): comp is not a grammatical function. She accounts for all previous uses of comp in other ways, and shows that a comp-less theory captures more generalizations. Extra morphological evidence for this view: verbs that allow extraction (bridge verbs) are those which agree in definiteness with their clausal complements; those complements are obj. If the clausal argument of a bridge verb is obj, then the focus-raised accusative can t be. { } mond-t-ák (26) a. A fiú-k, [hogy vár-j-ák Évát] { *mond-t-ak } say-past-3pl.def the boy-pl comp wait-sbj-3pl.def Eve-acc say-past-3pl.indef The boys said that they expected Eve. { }?telefonál-t-ák b. A fiú-k, [hogy vár-j-ák Évát] { telefonál-t-ak } telephone-past-3pl.def the boy-pl comp wait-sbj-3pl.def Eve-acc telephone-past-3pl.indef The boys telephoned that they expected Eve. { } *dicseked-t-ék c. A fiú-k, [hogy vár-j-ák Évát] { dicseked-t-ek } brag-past-3pl.def the boy-pl comp wait-sbj-3pl.def Eve-acc brag-past-3pl.indef The boys were bragging that they expected Eve. (27) a. A fiú-k Évá-t i mond-t-ák, hogy vár-j-ák t i the boy-pl.nom Eve-acc say-past-3pl.def comp wait-sbj-3pl.def The boys said that they expected Eve. b.? A fiú-k Évá-t i telefonál-t-ák, hogy vár-j-ák t i the boy-pl.nom Eve-acc telephone-past-3pl.def comp wait-sbj-3pl.def The boys telephoned that they expected Eve. c. * A fiú-k Évá-t i disceked-t-ék, hogy vár-j-ák t i the boy-pl.nom Eve-acc brag-past-3pl.def comp wait-sbj-3pl.def The boys were bragging that they expected Eve. Conclusion: a verb can agree in definiteness with an NP that is not its object. 5
3 Accusative case Does a verb just agree in definiteness with anything nearby that has accusative case? Instead of: (28) -Vm V infl ( obj def= c +) (29) -Vk V infl ( obj def) How about: (30) -Vm V infl ( a case)=acc ( a def)= c + (31) -Vk V infl ( a case)=acc ( a def) Counterexample 1 Indefinite accusative temporal adverbial, definite V: (32) Két orát tanul-t-a a lecké-t two hour-acc study-past-3sg.def the lesson-acc I studied the lesson (for) two hours. Counterexample 2 Accusative focus-raised objects optionally agree in definiteness with matrix V: (33) a. Két lány-t mond-t-ál hogy Zoli fel-hiv-ott two girl-acc say-past-2sg.indef comp Zoli.nom up-call-past.3sg.indef You said that Zoli invited two girls. b. Két lány-t mond-t-ad hogy Zoli fel-hiv-ott two girl-acc say-past-2sg.def comp Zoli.nom up-call-past.3sg.indef You said that Zoli invited two girls. Counterexample 3 Definite focus-raised objects may occur with indefinite matrix V (Szamosi, 1976): (34) a. A könyv-et kér-t meg, [hogy hoz-z-am el]. the book-acc ask-past.3sg.indef perf comp bring-sbj-1sg.def away It was the book that he asked me to bring. b. A könyv-et kér-te meg, [hogy hoz-z-am el]. the book-acc ask-past.3sg.def perf comp bring-sbj-1sg.def away It was the book that he asked me to bring. 6
4 Solution Definiteness agreement is enforced under accusative case assignment. Temporal adverbials license own case don t agree with local V. Focus-raised objects can get case from downstairs V or upstairs V optionally agree with matrix V. Focus-raised subjects must get case from matrix V obligatorily agree with matrix V. Accusative case assignment satisfies two properties: Locality: acc is assigned to the local f-structure (via single-attribute variable a). Complement-directedness: acc is assigned towards complements (via comppath). (35) comppath=obj + (subj) Lexical specifications: (36) Definite verbs: The GF whose accusative case feature I assign must be definite : ( a case)=acc ( a)= c ( comppath) ( a def)= c + (37) Indefinite verbs: If I assign you accusative case, you better not be definite : { ( a case)=acc ( a)= c ( comppath) ( a def) } 5 Refinement: Infinitive-selecting verbs Some infinitive-selecting verbs object-agree with object of infinitive complement (xcomp): (38) a. Meg-próbál-ok ritk-ább-an veszíteni el dolg-ok-at. perf-try-1sg.indef rare-more-ly lose.inf away thing-pl-acc I m trying to lose things more rarely. b. Meg-próbál-om ritk-ább-an veszíteni el az esernyőm-et. perf-try-1sg.def rare-more-ly lose.inf away the umbrella-acc I m trying to lose my umbrella more rarely. Slight problem: constraints on acc assignment are too rigid; there is no single-attribute path from matrix verb to embedded object, if they are separated by xcomp. 7
New accusative case assignment conditions: Locality: acc is assigned within the minimal finite domain. Complement-directedness: acc is assigned to complements (and xcomp counts). 6 Remaining questions 1. Why do subjects change case when they focus-raise? 2. Why is case switch obligatory for some speakers, optional for others (Gervain, 2002)? 3. Why does definiteness agreement have a different distribution from person agreement in Hungarian? (Not inconsistent with the analysis, but still an open question.) 4. How does this notion fare in other languages with object agreement? 5. How general is the connection between case and agreement? (Icelandic: subject agreement is with nom argument.) References Berman, J. (2003). Clausal Syntax of German. CSLI Publications. Coppock, E. (2003). Sometimes it s hard to be Coherent. In Procedings of LFG03. den Dikken, M. (1999). Agreement and clause union. Ms., CUNY graduate center. Farkas, D. and de Swart, H. (2003). The Semantics of Incorporation. CSLI Publications. Gervain, J. (2002). Linguistic methodology and microvariation in language: the case of operator-raising in Hungarian. Master s thesis, University of Szeged, Hungary. Mithun, M. (1984). The evolution of noun incorporation. Language, 60:847 93. Szamosi, M. (1976). On a surface structure constraint in Hungarian. In McCawley, J., editor, Syntax and Semantics, volume 7, pages 409 426. 8